PartitionAndSerialize Operator: Allocate keyOutputBuffer with pre-determined size

[emilysun201309]

Summary:
Pre-determine the output buffer size for sort keys by applying BinarySortableSerializer::serializedSizeInBytes on the input data.
Use both the valueOutputBufferSize and keyOutputBufferSize to determine batch size
Allocate the buffer size needed and avoid resizing the buffer

Reviewed By: xiaoxmeng

Differential Revision: D76008316

== RELEASE NOTES ==
Prestissimo (Native Execution) Changes
* Improve memory usage in PartitionAndSerialize Operator by pre-determine the serialized byte size of a given sort key at 'rowId'. This allows PartitionAndSerialize Operator to pre-allocated the exact output buffer size needed for serialization, avoid wasted memory allocation. User should expect lower memory usage and up to 20% runtime increase when serializing a sort key. 

[facebook-github-bot]

This pull request was exported from Phabricator. Differential Revision: D76008316

[steveburnett]

Thanks for the release note entry!

• Please revise the entry to begin with a keyword in the Order of changes in the Release Notes Guidelines such as Fix or Improve.
• Can you describe this PR's effect from a Presto user's perspective? What would a Presto user notice from the change in this PR? (This may help identify how to address the first point.)

[yhwang]

Hi @emilysun201309, I am working on the release note of this PR. Can you summarize the note and use user-understandable wording? Thanks!